AI Paper Atlas

Research map / Detection and segmentation / 3D detection and driving

DeepInteraction++: Multi-Modality Interaction for Autonomous Driving

Zeyu Yang, Nan Song, Wei Li, Xiatian Zhu, Li Zhang, Philip H. S. Torr

TPAMI 2025 · Detection and segmentation › 3D detection and driving · approach: LiDAR perception for driving · 28 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add image-to-LiDAR representational interaction with L2I (bilateral instead of unilateral) (ablation)DeepInteraction · nuScenes valmAP: 69.2 -> 69.9 with image in decoder; 68.9 -> 69.4Improved
add image representation in decoder (MMPI on both modalities) (ablation)DeepInteraction · nuScenes valmAP/NDS: 69.2 -> 69.9 mAP, 72.2 -> 72.6 NDSImproved
add multi-modal representational interaction encoder (ablation)DeepInteraction++ · nuScenes valmAP: 67.7 -> 70.6Improved
add LiDAR-guided polar attention (ablation)DeepInteraction++ · nuScenes valmAP/NDS: 70.32 -> 70.63 mAP, 73.02 -> 73.27 NDSImproved
add LiDAR-guided polar attention (ablation)DeepInteraction++ · Waymo (1/5 train)L2 mAP: 66.95 -> 67.02No clear effect
keep image features in perspective form instead of BEV (ablation)DeepInteraction++ · nuScenes valmAP: 69.4 -> 70.3Improved
replace DETR decoder layers with MMPI for both modalities (ablation)DeepInteraction++ · nuScenes valmAP: 69.7 -> 70.6Improved
use alternating LiDAR and image interaction in decoder (ablation)DeepInteraction++ · nuScenes valNDS: 72.8 -> 73.3Improved
add multimodal interaction to the LiDAR-only baseline (pillar backbone)DeepInteraction++ · nuScenes val (pillar)mAP: 54.5 -> 65.6Improved
vary number of queries at train/inference (200/300/400) (ablation)DeepInteraction · nuScenes valmAP: 69.7 to 70.1No clear effect
swap DeepInteraction for DeepInteraction++ with R50 backboneDeepInteraction · nuScenes valNDS: 72.6 -> 72.9Improved
upgrade DeepInteraction to DeepInteraction++ for end-to-end planningDeepInteraction · nuScenes val planningavg collision rate (%): 0.25 -> 0.19Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy